v1.2.1

v1.2.1: Local AI (Ollama) & QoL Changes.

What's New?

Native Ollama Support

You are no longer dependent solely on Gemini. It is now possible to run content-summarizer 100% locally with Ollama.

• Added --provider ollama flag.
• New configuration options: --ollama-model, --ollama-url, --ollama-ctx.

content-summarizer summarize "URL" -p ollama --ollama-model <model_name>

Instant Startup

We rewrote how heavy dependencies (like pytorch, pytubefix, and faster-whisper) are loaded.

• Drastically reduced boot time: From ~0.8s down to < 0.3s.
• The CLI now responds instantly for commands like --help or initial validations.

Whisper Unicode Crash Fix

Fixed a bug where faster-whisper (via PyAV/FFmpeg) would crash upon encountering special characters on systems configured as pure ASCII.

Organized CLI Help

The --help command has been fully refactored. Flags are now logically grouped for better readability.

New Gemini Models

Updated the list of supported models to include Google's latest versions.

Deprecations

Remote API (--api): The functionality to offload transcription to a remote API is now deprecated and will be removed in v2.0.0.

How to Upgrade?

Bash

uv tool upgrade content-summarizer

Full Changelog: v1.1.0...v1.2.1

v1.1.1 - Hotfix

What's New?!

This is a small patch following the major v1.1.0 release.

Improvement

• Log file path is now show when an error occurs.

v1.1.0

What's New?!

New flag: --compute-type

• Specify the compute type for local transcription (Faster-Whisper).
• Accepted values: int8, float16, float32, int8_float16 and auto.
• Default value if the flag is not passed: auto.
• Works in both summarize and config subcommands.

Migrated to the new Gemini SDK

• Content-Summarizer now uses the new Gemini SDK.
• The list of accepted models changed: 2.5-flash, 2.5-pro, 3-flash and 3-pro.
• Default model if --gemini-model or -g is not passed: 3-flash.

Bug Fix

• Fixed a crash related to locale/encoding on some Linux distributions (UnicodeDecodeError).

v1.0.4

What's new?!

• Quicker initialization, you no longer have to wait 1+ seconds to use an content-summarizer summarize -h.
• New command! content-summarizer -v will show the current version! You can also use content-summarizer --version.
• Some README updates.